I think it is highly cited because it deals with an important and topical issue—the question of whether Earth’s troposphere has or has not warmed over the past two-and-a-half decades.

Since 1979, Microwave Sounding Units (MSUs) flown on polar-orbiting satellites have been used to monitor the temperatures of broad layers of the atmosphere. In the early 1990s, scientists at the University of Alabama at Huntsville (UAH) attempted to generate a homogeneous record of multi-decadal atmospheric temperature change by splicing together the MSU temperatures from over a dozen satellites. The UAH results initially showed cooling of the lower atmosphere from 1979 to the present, particularly in the tropics.

“Several prominent politicians have used the lack of warming in the UAH TLT data to argue that the Earth is not warming, and that all climate models are fundamentally flawed. Our work suggests that such criticisms are misguided and ill-informed.”

This cooling was at variance with surface thermometer records, which indicated that the surface of the Earth had warmed since 1979. The UAH MSU record was also in fundamental disagreement with both climate models and with basic theory, which predicted that the lower atmosphere should warm in response to human-caused increases in greenhouse gases. This left the scientific community with a real dilemma: Why couldn’t we see tropospheric warming in MSU data? And how could the Earth’s surface be warming, but the air above it be cooling?

Our paper describes a synthesis of knowledge rather than a new discovery or methodology. The synthesis involved information from many different climate models, from multiple observational datasets, and from basic theory.

Our paper compared the relationship between surface and tropospheric temperature changes over a wide range of observational and climate model datasets. These comparisons were performed on multiple timescales, using month-to-month, year-to-year, and decade-to-decade temperature changes.

We focused on the deep tropics, where different research groups obtained quite different estimates of "observed" tropospheric temperature trends. Our intent was to see whether we could constrain these uncertainties using knowledge of the basic physics that controls the vertical temperature structure in the tropical atmosphere.

One of the key aspects of this "basic physics" involves the release of latent heat when moist, convecting air rises and condenses to form clouds. Because of latent heat release, tropical temperature changes averaged over large areas—and averaged over enough time to damp day-to-day "weather noise"—are generally larger in the lower and mid-troposphere than at the surface. This "amplification" behavior was well-known from theory and from climate model results.

We found rather puzzling amplification results for the UAH lower tropospheric temperatures (TLT). For "fast" (month-to-month and year-to-year) fluctuations in tropical temperatures, UAH TLT anomalies were 1.3 to 1.4 times larger than surface temperature anomalies, consistent with models, theory, and other observational datasets. But for "slow" (decade-to-decade) temperature changes, the UAH TLT trends were noticeably smaller than surface trends.

In contrast, amplification results from nearly two dozen computer models were consistent with theory across all timescales considered, despite large differences in model structure, physics, and climate forcings.

We also saw tropospheric amplification of surface temperature changes in a second observational TLT dataset developed by the Remote Sensing Systems group (RSS) in California. Although RSS and UAH scientists relied on the same raw MSU data, they made different decisions on how to adjust that data for the effects of drifts in satellite orbits and for instrument calibration problems.

One possible explanation for our results is that the UAH data are reliable, and that different physical mechanisms control the response of the tropical atmosphere to "fast" and "slow" surface temperature fluctuations. If so, all models must have common errors that prevent them from capturing these different physical mechanisms, which have yet to be identified.

A second explanation is that significant inhomogeneities remain in the UAH tropospheric temperature records, leading to residual cooling biases in the UAH long-term trend estimates. In our view, this second explanation is simpler and more plausible, given the consistency of amplification results across models and timescales, our theoretical understanding of how the tropical atmosphere should respond to sustained surface heating, and the currently large uncertainties in observed tropospheric temperature trends.

My own involvement in this research was prompted by pervasive claims that MSU data constituted unambiguous "proof" that climate models were unreliable and the surface thermometer record was wrong. I felt such claims were not justified by the available scientific evidence, and that uncertainties in satellite-based estimates of tropospheric temperature change were much larger than some had contended.

This research did face certain obstacles along the way, in part because of the not inconsiderable political attention that the "great MSU debate" had received.

Several prominent politicians have used the lack of warming in the UAH TLT data to argue that the Earth is not warming, and that all climate models are fundamentally flawed. Our work suggests that such criticisms are misguided and ill-informed.

The elegant research of Carl Mears and Frank Wentz (at RSS) and Steve Sherwood, John Lanzante, and Cathryn Meyer (at Yale University and at the Geophysical Fluid Dynamics Lab in Princeton) clearly demonstrates that there are significant uncertainties in estimates of "observed" tropospheric temperature changes obtained from satellites and weather balloons.

Our best current understanding of the observational record is that the tropical troposphere has warmed—and not cooled—since 1979. This warming is broadly consistent with results from climate models and basic theory.

We have found that, like beauty, model-versus-data agreement very much depends on one’s observational perspective. In my opinion, these findings enhance confidence in the usefulness of computer models for studying future climate change.